Volume 118, 2016, pp. 513–531 DOI: 10.1650/CONDOR-15-168.1 RESEARCH ARTICLE Influence of landscape, habitat, and species co-occurrence on occupancy dynamics of Canada Warblers

Alexis R. Grinde1* and Gerald J. Niemi1,2

1 Natural Resources Research Institute, University of Minnesota Duluth, Duluth, Minnesota, USA 2 Departments of Biology and Integrated Biosciences, University of Minnesota Duluth, Duluth, Minnesota, USA * Corresponding author: [email protected] Submitted September 30, 2015; Accepted May 2, 2016; Published June 29, 2016

ABSTRACT The Canada Warbler ( canadensis) is a species of high conservation importance because of its low overall density and long-term widespread population declines on the breeding grounds. Results of previous research on the species suggest that its breeding habitat preferences vary across its range. However, the underlying processes associated with habitat use are unknown. Using a 20 yr dataset, we developed occupancy models for Canada Warblers to determine the influence of habitat characteristics (e.g., understory vegetation, canopy cover), landscape context (e.g., edge, forest patch size), and species co-occurrence on occupancy, colonization, and local extinction parameters. Our results show that multiple habitats are used by Canada Warblers on the breeding grounds; common variables associated with large-scale, long-term occupancy dynamics are forest age, landscape composition at the 100 m and 500 m scales, and mean patch size. Overall, Canada Warblers were nearly twice as persistent in mature forest stands (.80 yr) and large, mixed forest stands. Further, models indicated that species co-occurrence was an important predictor of Canada Warbler occupancy in some cover types. The results of this study increase our understanding of population processes over large, dynamic landscapes and provide essential conservation information to improve habitat and landscape management for the Canada Warbler. Keywords: avian conservation, Canada Warbler, competition, forest management mechanisms, habitat use, landscape, occupancy

Influencia del paisaje, del ha´bitat y de la presencia simulta´nea de especies en la dina´mica de ocupacion´ de Cardellina canadensis RESUMEN Cardellina canadensis es una especie de alta importancia de conservacion´ debido a su baja densidad global y a la disminucion´ poblacional generalizada de largo plazo en los sitios reproductivos. Investigaciones previas sobre la especie sugieren que las preferencias de ha´bitat reproductivo var´ıan a lo largo de su rango. Sin embargo, los procesos subyacentes asociados con el uso de ha´bitat son desconocidos. Usando una base de datos de 20 anos,˜ desarrollamos modelos de ocupacion´ para C. canadensis para determinar la influencia de las caracter´ısticas del ha´bitat (e.g., vegetacion´ del sotobosque, cobertura del dosel), el contexto del paisaje (e.g., borde, tamano˜ del parche de bosque) y la presencia simulta´nea de especies en la ocupacion,´ colonizacion´ y para´metros locales de extincion.´ Nuestros resultados muestran que C. canadensis usa multiples´ ha´bitats en los sitios reproductivos; las variables usuales asociadas con la dina´mica de ocupacion´ de gran escala y largo plazo son la edad del bosque, la composicion´ del paisaje a la escala de 100 m y 500 m, y el tamano˜ promedio del parche. En general, los individuos de C. canadensis fueron casi el doble de persistentes en las masas de bosque maduro (.80 anos)˜ y en las grandes masas boscosas mixtas. Ma´saun,´ los modelos indicaron que la presencia simulta´nea de especies fue un predictor importante de la ocupacion´ de C. canadensis en algunos tipos de cobertura. Los resultados de este estudio aumentan nuestro entendimiento de los procesos poblacionales a lo largo de grandes paisajes dina´micos y brindan informacion´ esencial de conservacion´ para mejorar el manejo del ha´bitat y del paisaje para C. canadensis. Palabras clave: Cardellina canadensis, competencia, conservacion´ de aves, mecanismos de manejo del bosque, ocupacion,´ paisaje, uso de ha´bitat

INTRODUCTION Canada, the northeastern United States, and south along the to Tennessee (Reitsma et al. The Canada Warbler (Cardellina canadensis) is a Neo- 2010). Canada Warblers winter in northern South Amer- tropical migratory that breeds in forests of boreal ica, with the greatest numbers observed in and east of the

Q 2016 Cooper Ornithological Society. ISSN 0010-5422, electronic ISSN 1938-5129 Direct all requests to reproduce journal content to the Central Ornithology Publication Office at [email protected] 514 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi

Andes (Reitsma et al. 2010). The species’ breeding habitat Species at Risk Act (COSEWIC 2012, Environment has been suggested to vary across its range (Reitsma et al. Canada 2015). In the United States, it is included on the 2010). Canada Warblers inhabit a wide range of deciduous list of species of conservation concern at the national level and coniferous forests but are most common in mixed (U.S. Fish and Wildlife Service 2008) and is considered a coniferous–deciduous forests that are moist and have well- ‘‘Yellow Watch List’’ species (Butcher et al. 2007). developed understory shrub layers (Reitsma et al. 2010). In Population trend estimates for Canada Warblers in the the eastern portion of its range the species breeds both in boreal hardwood transition zone ( Conservation early-successional forest stands and in deciduous and Region 12) have declined 2.85% since 1966. However, mixed-species understories of mature forests (DeGraaf et BBS population trends have been relatively stable in al. 1998, Reitsma et al. 2008). It appears to be disturbance Minnesota, USA (Sauer et al. 2014). dependent at higher elevations, especially in the southern Long-term and wide-ranging declines in Canada War- portion of its range (Lambert and Faccio 2005, Becker et al. bler populations suggest the need for additional informa- 2012). Quaking aspen (Populus tremuloides) and balsam tion on long-term occupancy dynamics. The Minnesota poplar (P. balsamifera) stands are favored in the central National Forest Breeding Bird Monitoring Program portion of the range, while mixed forests and forested (hereafter ‘‘Minnesota monitoring program’’) was estab- wetlands and swamps seem to be preferred in the northern lished in 1995, in response to concerns about biodiversity part (Schmiegelow et al. 2014, Niemi et al. 2016). and population declines of migratory , including Several fine-scale habitat factors are associated with the Canada Warbler (Hanowski and Niemi 1995, Hanowski Canada Warbler abundance during the breeding season. et al. 2005, Niemi et al. 2016). The program was designed The species inhabits many forest-cover types but is most to provide an estimate of population change for forest bird abundant in moist mixed forests with a dense understory species in national forests in Minnesota (Figure 1). Results and complex ground cover (Conway 1999, Becker et al. from the Minnesota monitoring program indicate stable 2012). Hallworth et al. (2008a) showed shrub stem density population trends for Canada Warblers (1995–2014: to be critical to breeding Canada Warblers and suggested annual change ¼ 0.53%, P ¼ 0.33) in Superior National that a dense understory enhances Canada Warbler Forest and significantly increasing trends (1995–2014: breeding habitat. Similarly, Goodnow and Reitsma (2011) annual change ¼ 4.87%, P ¼ 0.02) in Chippewa National found that forests with complex ground structure and Forest. Data from this monitoring program provide an thickets of small-stemmed woody plants increase Canada opportunity to assess habitat and landscape factors that Warbler nesting success. Although Canada Warblers are support stable or increasing Canada Warbler populations, considered sensitive to forest fragmentation (Enser 1992, using a long-term (20 yr) occupancy dataset. Freemark and Collins 1992), populations are known to In addition to habitat factors, species interactions have occupy young, disturbed forests in Wisconsin (Sodhi and long been thought to influence avian population dynamics Paszkowski 1995); areas that were heavily logged 5–15 yr (MacArthur 1958, Lovette and Hochachka 2006, Niven et previously in northern hardwood forests of al. 2009). Behavioral decisions based on both positive and (Webb et al. 1977); and areas heavily logged 15–25 yr negative interspecific interactions can influence important before in mixed forests of New Hampshire (Hallworth et aspects of a species’ ecology such as habitat selection, al. 2008a). The most important features in recently foraging, spatial distribution, and dispersal, and thus can harvested stands appear to be high understory shrub influence occupancy dynamics. The results of recent density (Schmiegelow et al. 1997, Hallworth et al. 2008b) studies (Gotelli et al. 2010, Ricklefs 2013) have shown and retention of residual trees (Hagan et al. 1997, Lambert that both aggregated and segregated distribution patterns and Faccio 2005, Hallworth et al. 2008b). among ecologically similar species result from a combina- Results from the North American Breeding Bird Survey tion of competitive and positive interactions, independent (BBS) indicate 4 decades of Canada Warbler decline of habitat characteristics, at local and broad spatial scales. throughout the northeastern breeding range, with annual We also used null models to investigate potential patterns population changes between 3.8% and 7.3% since 1980 of co-occurrence in bird community assemblages (Gotelli (Sauer et al. 2014). The reasons for the declines are 2001), in order to assess the influence of species co- unknown, but factors may include loss of wintering habitat occurrence on Canada Warbler occupancy in Minnesota’s in South America, loss and degradation of habitat on the national forests. breeding grounds (Lambert and Faccio 2005), and Our objective was to better understand factors associ- collisions with man-made structures (Loss et al. 2014, ated with long-term Canada Warbler occupancy dynamics Machtans and Thogmartin 2014). The species was over large, dynamic landscapes. We identified potential designated as ‘‘threatened’’ by the Committee on the Status variables important to Canada Warbler population dynam- of Endangered Wildlife in Canada (COSEWIC) in 2008, ics from the existing literature (e.g., Webb et al. 1977, and since 2010 has been listed as ‘‘threatened’’ under the Freemark and Collins 1992, Sodhi and Paszkowski 1995,

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 515

FIGURE 1. Stand locations of Minnesota National Forest Breeding Bird Monitoring Program point counts. Each point represents 3–5 forest stands (A) in northern Minnesota’s Chippewa (B) and Superior (C) National Forests (NF).

Schmiegelow et al. 1997, Conway 1999, Lambert and types in Superior and Chippewa are aspen (Populus spp.), Faccio 2005, Hallworth et al. 2008a, 2008b, Becker et al. paper birch (Betula papyrifera), (Picea spp.), 2012). We used habitat and landscape variables from the balsam fir (), tamarack (Larix laricina), literature along with species co-occurrence variables to and pine (Pinus spp.) forests. Superior and Chippewa have create candidate model sets to assess the influence of relatively similar proportions of upland deciduous (~35%) stand, landscape, and species co-occurrence on Canada and lowland (~25%) cover types. However, Warbler occupancy dynamics in 2 national forests. Specific Chippewa has more upland coniferous forest (~35%), objectives of this study were to identify (1) land-cover and primarily pine and spruce plantations, than Superior stand-level characteristics associated with colonization and (~20%), while Superior has more upland mixed forest local extinction probabilities in each national forest and (2) (i.e. aspen–fir; 17%) than Chippewa (~2%). Both national the influence of stand-level characteristics and species co- forests have relatively low proportions of lowland hard- occurrence on occupancy and local extinction probabilities wood cover. The mosaic of habitats found in this region in different cover types in each national forest. supports ~155 breeding species of forest-dwelling (Green 1995). The breeding bird communities of these METHODS hemiboreal forests are among the most diverse in North America (Niemi et al. 1998). Study Area The study was conducted from 1995 to 2014 in 2 national Sampling forests in Minnesota. Chippewa National Forest (hereafter At the onset of the Minnesota monitoring program, avian ‘‘Chippewa’’) encompasses .269,000 ha in north-central point-count sampling locations were distributed across the Minnesota (Figure 1) and includes approximately 1,300 forest mosaic in a stratified random manner (Figure 1; lakes and ponds, 1,500 km of running water, and 178,000 Hanowski and Niemi 1995). For each national forest, ha of wetlands. Superior National Forest (hereafter stands were selected so that the final proportion of stands ‘‘Superior’’) comprises .1,580,000 ha (Figure 1), a signif- of each forest-cover type was equal to the proportion of icant portion of the northeastern ‘‘arrowhead’’ region of forested land available. The sample of stands is therefore Minnesota, and includes .180,000 ha of open water with representative of the percent forest cover found in each .2,000 lakes and .3,500 km of streams. national forest. Selected stands were large enough to Superior and Chippewa are located near the ecotone of accommodate 3 replicate sampling points separated by a boreal and northern temperate forests and contain a mix minimum of 220 m (Figure 2). Point-count sampling in the of forest-cover types. The most representative forest-cover Minnesota monitoring program followed national and

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 516 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi

FIGURE 2. Land cover and patch configuration of 3 aspen stands, each consisting of 3 point counts. Stands represent examples of (A) high, (B) medium, and (C) low mean patch size and (A) low, (B) medium, and (C) high percent open cover in the 100 m and 500 m buffers around the stand. regional standards (Ralph et al. 1995, Howe et al. 1997). (estimated visually at 10% intervals), and shrub (diameter Ten-minute point counts were conducted at each point at breast height ,2.5 cm) density (on a scale of 1–5; between June and early July (Hanowski et al. 2005, Pattison et al. 2011) were recorded in the field at intervals Etterson et al. 2009, Niemi et al. 2016). Point counts were of 3–5 yr. Although the study area is extensively forested, conducted by trained observers from ~0.5 hr before to 4 we included distance to the nearest road for each stand; hr after sunrise on days with little wind (,15 km hr1)and the average distance from a point in a stand to the road little or no precipitation. All birds heard or seen from the was used as the ‘‘distance to road’’ covariate value for the point were recorded, and distance was estimated as ,50 stand (Lapin et al. 2013). Ages of forest stands were m, 50–100 m, or .100 m (Howe et al. 1997). We excluded obtained from the Interagency Information Cooperative, from analyses birds that were .100 m from the point. which has integrated forest inventory data measured in the field by foresters from several forest-management groups Stand Characteristics and Land-Cover Variables within Minnesota, including county, state, and federal A total of 308 stands (924 points) were included in the agencies (Skally 2000, Niemi et al. 2016). All variables were analysis: 183 stands (549 points) in Superior and 125 averaged across the 3 sample points in each stand to stands (375 points) in Chippewa. From existing literature calculate annual average stand values for the study period. on the Canada Warbler (Sodhi and Paszkowski 1995, Canada Warblers are territorial during the breeding Schmiegelow et al. 1997, Conway 1999, Lambert and season, with reported territory size averaging 1 ha but Faccio 2005, Hallworth et al. 2008a, 2008b, Reitsma et al. ranging up to 3.3 ha or larger (Hallworth et al. 2008a, 2010, Becker et al. 2012), we established a set of candidate Environment Canada 2015). However, results from recent variables that we thought would influence occupancy. studies have shown that landscape factors at larger spatial Covariates were calculated over the 20 yr study period, scales were important drivers for warbler species (Streby et using a variety of data sources. Percent canopy cover al. 2012, Lapin et al. 2013). On the basis of this information (estimated visually at 10% intervals), percent ground cover and available land-cover data, we chose to calculate cover-

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 517 type and landscape variables at scales of 100 m (3.14 ha) possibility of ‘‘habitat checkerboards.’’ These can occur and 500 m (78.5 ha), using ArcGIS 10.1 (ESRI, Redlands, when species are associated with different abiotic features California, USA). Stands were analyzed by placing a 100 m of the sites and lead to more or less co-occurrence than and 500 m buffer around each point-count location and expected by chance (Gotelli 2001). Data were organized in taking the union to form one asymmetrical buffer (Figure a matrix of presence (1) and absence (0), in which each 2; Lapin et al. 2013). Percent land-cover type in each buffer species represents a row and each stand a column. The C- was calculated using National GAP Land Cover data, score index (Stone and Roberts 1990) was used as a which uses Landsat 1999–2001 imagery as the base for its quantitative co-occurrence index; the C-score is the models and provides land-cover data at a 30 3 30 m number of checkerboard units for all unique pairs of resolution (U.S. Geological Survey 2011). A combination of species. We used fixed-sum row and column constraints field observation data, Forest Inventory Analysis data, and and a ‘‘sequential swap’’ algorithm for randomizing Global Forest Change data (Hansen et al. 2013) was used matrices and ran 10,000 simulations (Ulrich and Gotelli to assess changes in land cover over the duration of the 2007a, 2007b, and Ulrich et al. 2009). Using fixed-sum row study, including the occurrence of logging within the and column constraints produces null matrices with the buffers during the study. We classified land-cover types same number of site co-occurrences per species (row into 6 broad categories, including 4 forest-cover types: totals) and the same number of species per site (column lowland coniferous forest, upland coniferous forest, upland total) as observed in the original dataset. The sequential deciduous forest, and upland mixed forest. ‘‘Open’’ and swap algorithm reshuffles the original matrix by repeatedly ‘‘aquatic’’ were also included as land-cover categories. swapping submatrices that preserve row and column Land-cover types classified as aquatic are those that are totals, and it is not prone to Type I or Type II errors open water year round, such as lakes, open wetlands, (Gainsbury and Colli 2003, Gotelli and Entsminger 2005). ponds, and streams. Open cover types were considered habitats with ,10% canopy cover, generally clear-cut Occupancy Analyses areas, wetlands, agricultural land, pastures, and shrub- We used a single-species, multiseason occupancy modeling lands. The Patch Analyst 5 application for ArcGIS (Rempel framework (MacKenzie et al. 2002) to evaluate the et al. 2012) was used to calculate mean patch size (ha) and influence of stand characteristics, landscape, and species total edge (m) for the 100 m and 500 m buffers around the co-occurrence on Canada Warbler occupancy dynamics. 3 sample points of each stand in 5 yr intervals, to track Multiseason occupancy modeling allows spatial and changes in patch size and edge density over the course of temporal dynamics of populations to be modeled and the study (McGarigal and Marks 1995). Because of the analyzed and accounts for errors in detection during duration of the study and inconsistencies between datasets, sampling. These models use information (detection or we were not able to develop a precise, reliable measure for nondetection) from repeated observations (multiple points extent of logging or logging type (clear-cut, partial harvest) per stand), which are used to estimate probabilities of that occurred in 100 m and 500 m buffers around the occupancy and detectability (MacKenzie et al. 2006). Our stands. For this reason, we included logging as a binary occupancy analysis utilized spatial replicates (3 points variable. We recognize that this is not an ideal method for stand1) associated with the monitoring program’s exper- including logging disturbance in the models. However, imental design to estimate detection probabilities. Multiple logging disturbance was also indirectly incorporated at the studies (e.g., Kendall and White 2009, Guillera-Arroita stand level as a function of stand age and at the landscape 2011) have examined potential estimation bias associated level as a function of average percent open habitat over with spatial replication. We recognize the potential for bias time. when using spatial replicates, but we consider that the value of inference gained from applying occupancy models Co-occurrence Analysis to this long-term dataset outweighed the issues associated To generate a list of species that may influence Canada with biased estimations. Warbler occupancy, EcoSimR’s co-occurrence module Multiseason occupancy models assume that the popu- (Gotelli and Ellison 2013) was used to assess patterns of lation is open between seasons or that stands that were species co-occurrence. The co-occurrence module tests for previously occupied can become unoccupied, and stands nonrandom patterns of co-occurrence in a presence– that were previously unoccupied can become occupied, absence matrix by using Monte Carlo randomizations that between seasons. In the occupancy modeling framework, create ‘‘pseudo-communities’’ (Pianka 1986), then statisti- estimates associated with these processes are referred to as cally compares the patterns in these randomized commu- probability of local extinction (i.e. site abandonment) and nities with those in the real data matrix (Diamond 1975). probability of colonization. Multiseason occupancy mod- Species richness for each stand was analyzed for each eling considers the dynamic changes in occupancy as a national forest and by cover type to minimize the first-order Markov process (MacKenzie et al. 2006) where

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 518 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi the probability of a site being occupied in season t þ 1 dynamics in each cover type. For example, the importance depends only on the occupancy state of the site in the of ground cover may be similar in upland deciduous and previous season, t. Modeling changes in occupancy as a upland conifer stands but may not have the same influence Markov process accounts for temporal autocorrelation in lowland conifer stands. Models were subset into ‘‘stand- (MacKenzie et al. 2006) when observations on the same characteristic variables’’ (age, ground cover, shrub density, sampling unit are positively correlated. Parameter esti- distance to road, logging disturbance, mean patch size mates were related to covariates using the logistic equation within the buffer, total edge within the buffer, percent open or logit-link function. This general framework was used to habitat in 500 m buffer, and percent aquatic habitat in 500 develop a set of candidate models that address competing m buffer ) and ‘‘species co-occurrence variables’’ (abun- hypotheses of occupancy dynamics. We ranked models on dance of potentially competing or positively interacting the basis of AICc (Akaike’s Information Criterion corrected species, based on co-occurrence results; Appendix Table for sample size), and the strength of evidence for each 2). Occupancy models for this analysis estimated the model was determined using DAICc (Burnham and influence of covariates on w and e and also estimated p Anderson 2002). All analyses were implemented in (Appendix Table 2; MacKenzie et al. 2006). PRESENCE 6.9 (Hines 2006) to obtain maximum-likeli- We used the same model-building approach for both hood parameter estimates. Preliminary analyses indicated analyses. First, we evaluated single predictor models and that Canada Warbler occupancy dynamics differed be- determined predictors with the highest model weight(s) tween Chippewa and Superior; moreover, the dynamics based on AICc from each variable subset. The best differed between cover types within each forest. Therefore, predictors (wi . 5%) were added to build a ‘‘full model’’; we used 2 approaches to model Canada Warbler occupan- predictors that were nonsignificant in the full model were cy dynamics. removed to produce final occupancy models. If the final Forest dynamics. Candidatemodelsforthefirst models included more than one predictor, we evaluated modeling approach were built to determine the land-cover variables for collinearity and interactions. This procedure and stand-level characteristics influencing occupancy allowed us to minimize the number of models examined dynamics in each national forest. For this analysis, national and avoid collinearity issues. Initial modeling of observer forests were analyzed separately and all stands within each effects and habitat on p resulted in model convergence forest were included in the analysis. Models were built issues. For this reason, we chose to model detection using stand-level and land-cover covariates that were probability as a function of year to represent differences in calculated at the 100 m and 500 m buffer scales. At each observers and habitats between seasons. We modeled p as scale, we subset the variables into ‘‘stand-characteristic a constant (without a covariate relationship) and as a variables’’ (age, ground cover, shrub density, distance to function of year (Appendix Table 1); and, using the most road, logging disturbance, mean patch size within the general model structure for occupancy dynamics, we buffer, and total edge within the buffer) and ‘‘land-cover selected the covariance structure on p with the lowest variables’’ (percent upland conifer, upland deciduous, AICc value. We then fit the remaining occupancy models upland mixed, lowland conifer, open, and aquatic cover separately for each candidate model subset (Appendix within the buffer; Appendix Table 1). Tables 1 and 2). We used the parametric bootstrap Occupancy models estimated the probabilities of initial procedure for assessing the goodness-of-fit of our best site occupancy (w), colonization (c), local extinction (e), model (MacKenzie and Bailey 2004). Results are presented and detection (p) (Appendix Table 1; MacKenzie et al. as means 6 SE. 2006). Here, colonization was defined as the event of a site being occupied during survey t when it had not been RESULTS during survey t 1. By contrast, local extinction was the event of a site being unoccupied during survey t when it Observation Summary had been occupied at survey t 1. This type of model A total of 2,163 Canada Warblers were observed within a consists of a probability of occupancy estimated for the 100 m radius during 1995–2014 in a total of 243 stands. first survey’s w0, whereas subsequent occupancy states The majority of these detections (1,844) occurred in depend on the probability of colonization and local Superior in a total of 168 stands; another 319 were extinction between consecutive survey years (MacKenzie recorded in a total of 75 stands in Chippewa. Over the 20 et al. 2006). We chose to model initial occupancy without yr of the study the average naive stand occupancy rates covariates to focus investigation on rate parameters for this were 0.28 (range: 0.15–0.37) and 0.08 (range: 0.05–0.14) in portion of the analysis (after Ferraz et al. 2007). Superior and Chippewa, respectively. Average observed Cover-type dynamics. This portion of the analysis frequency of stand occupancy (i.e. how frequently a stand considered stands by cover type, in order to determine was occupied) over 20 yr was 6.12 (range: 1–17) in similarities and differences in drivers of occupancy Superior and 5.10 (range: 1–17) in Chippewa. Observed

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 519 mean abundance of Canada Warblers in occupied stands Cover-type dynamics. Three land-cover types were was 2.22 birds stand1 (range: 1.5–2.10) in Superior and analyzed in Chippewa: upland deciduous, upland conifer, 1.65 birds stand1 (range: 1.1–2.29) in Chippewa. Detec- and lowland conifer. An additional habitat type, mixed tions varied by forest-cover type in both forests. In forest, was analyzed in Superior but excluded from Superior, detections were highest in mixed forest (0.78 Chippewa because of low representation. The best models detections stand1), lowland conifer (0.70 detections for Chippewa indicate that age, percent open habitat in the stand1), and upland deciduous (0.63 detections stand1) 500 m buffer, and abundance of Black-and-white Warblers cover types. In Chippewa, detections were highest in were significant covariates for occupancy and local extinc- lowland conifer (0.44 detections stand1) and upland tion probabilities. In upland deciduous stands, probability of conifer (0.12 detections stand1). occupancy was higher in younger stands (average age ¼ 41 yr; range: 5–89 yr). Probability of extinction decreased in Co-occurrence Models stands with a greater percent open habitat in the 500 m Co-occurrence analysis was conducted for each cover type buffer in upland deciduous stands (bage ¼0.59 6 0.09; in both national forests. Species that had high (ranked in bopen500 ¼0.48 6 0.19; Figure 5A, 5B). However, in top 5) checkerboard-unit values with Canada Warblers in lowland conifer stands, probability of occupancy decreased 3 forest-cover types were used as covariates in occupancy as the amount of open habitat in the 500 m buffer increased models; these values can represent positive or negative (bopen500 ¼0.34 6 0.19; Figure 5C). Black-and-white associations with Canada Warblers. Six species fit these Warbler abundance was negatively associated with Canada criteria: Black-and-white Warbler (n ¼ 3,983), Mourning Warbler occupancy in upland conifer stands (bBAWW ¼ Warbler (n ¼ 3,342), Warbler (n ¼ 2,611), Black- 1.60 6 0.32; Figure 5D) in Chippewa; no habitat covariates throated Green Warbler (n ¼ 3,644), Blackburnian Warbler were included in this cover type. (n ¼ 4,142), and Hermit Thrush (n ¼ 2,720) (see Appendix The best models in Superior indicated that age and Table 3). mean patch size were important variables for local extinction. Hermit Thrush and Magnolia Warbler abun- Occupancy Models dances also influenced occupancy. Specifically, younger Forest dynamics. The final model for Chippewa showed stands had a higher probability of local extinction in that age, percent upland conifer in the 100 m buffer, and upland deciduous (bage ¼0.31 6 0.15; Figure 5E), percent lowland conifer in the 500 m buffer were lowland conifer (bage ¼0.14 6 0.04; Figure 5G), and significant predictors of local extinction or site abandon- upland conifer (bage ¼0.34 6 0.16; Figure 5I) stands in ment. The Chippewa final model indicated that the Superior. Hermit Thrush abundance negatively affected estimated extinction coefficient was lower in older stands, the probability of Canada Warbler occupancy in upland indicating a greater rate of habitat abandonment in deciduous (bHETH ¼0.34 6 0.16; Figure 5F) and upland younger stands over time (bage ¼0.22 6 0.13; Figure conifer stands (bHETH ¼0.68 6 0.33; Figure 5J). In 3A). Stands with greater percent lowland conifer cover in lowland conifer stands, Magnolia Warbler abundance was the 500 m buffer (blowlandcon500 ¼0.64 6 0.17; Figure 3B) positively associated with the probability of occupancy and stands with greater percent upland conifer cover in the (bMAWA ¼ 0.45 6 0.16; Figure 5H). In mixed forest stands, 100 m buffer (buplandcon100 ¼0.28 6 0.11; Figure 3C) also local extinction probabilities were lower in stands with had lower probabilities of local extinction. The final model larger patches of mixed forest habitat at the 100 m scale indicated that detection probability in Chippewa was (bmps100 ¼1.96, SE ¼ 0.28; Figure 5K). constant over time. In Superior, stand age and percent upland deciduous DISCUSSION forest in the 500 m buffer around the stands were included in the final model as predictors of local extinction. Similar Our results indicate that Canada Warblers utilize a variety to the final model for Chippewa, the results of the final of cover types. Canada Warbler occupancy was highest in model showed that local extinction was less common in mixed forest and lowland conifer stands (Figure 6A, 6B). older stands (bage ¼0.13 6 0.003; Figure 3D) and that the These findings are consistent with the results of several probability of local extinction was lower in stands with a studies that reported the species to be associated with a greater amount of upland deciduous habitat in the 500 m variety of forest types, but to be most common in wet, buffer around the stand (buplanddec500 ¼0.08 6 0.002; mixed deciduous–coniferous forest with a well-developed Figure 3E). The top model in Superior indicated that shrub layer (Conway 1999, Reitsma et al. 2008, 2010, detection probability for Canada Warblers varied by year. Schmiegelow et al. 2014, Niemi et al. 2016). However, our We used estimates from best models for each forest to results provide inference about processes associated with derive average probability-of-occupancy estimates and differential habitat use that extends our understanding of standard deviations for each cover type (Figure 4). long-term habitat use over a large, dynamic landscape.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 520 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi

FIGURE 3. Predicted probabilities (6 SE) of local extinction for Canada Warblers, based on final models for forest dynamics in Chippewa (A–C) and Superior (D–E) National Forests, in relation to (A) natural log of stand age in Chippewa; (B) natural log of percent lowland cover in 500 m buffer in Chippewa; (C) natural log of upland conifer cover in 100 m in Chippewa; (D) natural log of stand age in Superior; and (E) percent upland deciduous cover in 500 m buffer around stands in Superior.

Forest Dynamics (.80 yr) forests with canopy gaps that promote a dense, Forest age significantly affected the probability of site well-developed shrub layer (Schieck et al. 1995, Enns and abandonment (local extinction) in both national forests, Siddle 1996, Cooper et al. 1997, Hobson and Bayne 2000b, indicating that the probability of local extinction is higher in Hobson et al. 2000, Schieck and Hobson 2000, Schieck et al. younger forests or that persistence is greater in mature 2000, Cumming and Machtans 2001, Machtans and Latour forest stands over time. These results were consistent across 2003, Hannon et al. 2004, Lambert and Faccio 2005). both national forests, even though the average age of Further, this species is positively associated with natural surveyed stands differed—the average age of forests in disturbances such as wind or tree-fall gaps (Hagan and Chippewa is 74 yr, compared to 112 yr in Superior. Grove 1999, Mitchell 1999, Faccio 2003) and outbreaks of Throughout their range, Canada Warblers breed in mature invasive insects such as the eastern spruce budworm

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 521

increases as habitat suitability decreases (Gates and Donald 2000, Broughton et al. 2013, Porter and Jarzyna 2013, Yntze et al. 2015); thus, the increased extinction estimates in early-successional stands suggest decreased habitat suitability for Canada Warblers, compared to mature stands. The increased probability of site abandonment in early-successional stands may be associated with posthar- vest stand characteristics (i.e. number of residual trees) or changes in stand characteristics that are associated with forest succession (i.e. shrub density). The results of several previous studies show the importance of shrub density and complexity of ground cover for Canada Warblers (Schmiegelow et al. 1997, Hallworth et al. 2008b, Becker et al. 2012). However, because of the spatial extent and duration of our study, detailed vegetation data were not consistently available for FIGURE 4. Predicted probabilities (6 SE) of occupancy by cover all stands. The importance of age in our models likely type for Canada Warblers, based on estimates from final models illustrates the interaction between stand age, succession, for Chippewa and Superior National Forests (NF). and associated changes to shrub canopy and ground cover. Long-term experimental studies that include detailed (Choristoneura fumiferana; Crawford and Jennings 1989), hydrological and vegetation assessments of early-succes- dynamics generally associated with mature forest stands. sional and mature stands will be beneficial for determining Our models suggest that older stands (.80 yr) provide the factors associated with stand abandonment that occurs stable long-term suitable habitat and thus have lower local as a result of the stand aging, compared with abandonment extinction probabilities. Zlonis and Niemi (2014) reported when stand characteristics remain unchanged. that Canada Warblers were more abundant in unmanaged The type of land cover important to Canada Warbler forests of the Boundary Waters Canoe Area Wilderness occupancy dynamics differed between forests, but land- (BWCAW) compared with managed stands in Superior. cover variables at the 500 m scale were included as Importantly, unmanaged stands in the BWCAW were, on covariates in the best models in both forests. This indicates average, older and more mixed; they had a greater variety in the importance of the larger landscape matrix to Canada shrub density and tree size classes than stands in Superior, Warbler dynamics. In Superior a greater amount of upland due to natural disturbances and gap dynamic processes deciduous habitat at the 500 m scale positively affected (Frelich and Reich 1995, Zlonis and Niemi 2014). persistence, and in Chippewa a greater amount of lowland Throughout its range, increases in local abundance of conifer at the 500 m scale and upland conifer at the 100 m Canada Warblers in regenerating forests (i.e. 5–30 yr scale had positive effects on species persistence. Results postdisturbance) following natural and anthropogenic from Superior are consistent with the finding of Schmie- disturbances have been reported (Titterington et al. 1979, gelow et al. (2014) that landscapes with a higher Christian et al. 1996, Hobson and Schieck 1999, Drapeau proportion of mixed wood and deciduous stands were et al. 2000, Hobson and Bayne 2000a, Schieck and Hobson more suitable for Canada Warblers. However, Schmiegelow 2000, Becker et al. 2012). Young forest stands likely provide et al.’s (2014) results also indicated that conifer stands were temporary open and shrub-like habitats with complex not selected by Canada Warblers across 3 bird conserva- ground cover that can increase Canada Warbler abundance tion regions in Canada. This difference in reported habitat and nesting success (Figure 6C; Schmiegelow et al. 1997, preference may be associated with the broader habitat Goodnow and Reitsma 2011). However, extent of harvest, classifications used by Schmiegelow et al. (2014) or may be harvest type, and number of residual trees that remain a reflection of differences in habitat availability between postharvest likely influence the suitability, and the duration the hemiboreal forests of Minnesota and the boreal forests of suitability, of postharvest stands for Canada Warblers. in the northern region of the species’ range. For example, Becker et al. (2012) found that Canada Warblers showed positive selection for 3 types of timber Cover-type Dynamics harvest (clear-cut, heavy partial, and light partial harvests), Age, percent open (nonforest) cover type at the 500 m yet response was greatest in sites with light partial harvest, scale, mean patch size at the 100 m scale, Black-and-white owing to the combination of a well-developed shrub story Warbler abundance, Hermit Thrush abundance, and and residual trees associated with this harvest type. Several Magnolia Warbler abundance were the variables associated studies have shown that the probability of local extinction with the best occupancy models in this portion of the

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 522 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi

FIGURE 5. Predicted probabilities (6 SE) of occupancy and local extinction for Canada Warblers, based on final models for cover- type dynamics in Chippewa (A–D) and Superior (E–K) National Forests, in relation to (A) natural log of stand age in upland deciduous stands; (B) natural log of percent lowland cover in 500 m buffer around upland deciduous stands; (C) natural log of percent open cover in 500 m buffer around lowland conifer stands; (D) natural log of Black-and-white Warbler (BAWW) abundance in upland conifer stands; (E) natural log of stand age in upland deciduous stands; (F) natural log of Hermit Thrush (HETH) abundance in upland deciduous stands; (G) natural log of stand age in lowland conifer stands; (H) natural log of Magnolia Warbler (MAWA) abundance in lowland conifer stands; (I) natural log of HETH abundance in upland deciduous stands; (J) natural log of HETH abundance in upland conifer stands; and (K) natural log of mean patch size in mixed forest stands. Abundance was measured as the number of birds heard or observed in 100 m buffer around stands.

FIGURE 6. Representative (A) mixed forest, (B) lowland conifer, and (C) early-successional cover types occupied by Canada Warblers.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 523 analysis. The relationship between age and the probability Species Co-occurrence of local extinction was similar to results of the forest-level Species co-occurrence may influence Canada Warbler occupancy model in lowland conifer stands in Chippewa occupancy. Our results indicated that in Superior, Mag- (average age ¼ 129 yr) and upland conifer stands in nolia Warbler abundance was positively associated with Superior (average age ¼ 89 yr); local extinction was lower Canada Warbler occupancy in lowland conifer stands, and in older stands. Hermit Thrush abundance was negatively associated with In upland deciduous stands in Chippewa and Superior, occupancy in upland conifer and upland deciduous stands. stand age was negatively associated with Canada Warbler In Chippewa, Black-and-white Warbler abundance was occupancy. The relationship between occupancy and stand negatively associated with occupancy of Canada Warblers age in upland deciduous stands seemingly contradicts the in upland conifer stands. Several recent studies have results reported above; however, these results speak to the shown that behavioral decisions based on species co- differential dynamics that occurred between cover types. occurrence influence habitat selection and persistence in The average ages of upland deciduous stands surveyed in many species and are a result of a combination of Chippewa (41 yr) and Superior (72 yr) were considerably competitive and positive interactions, often independent less than the average ages of all stands surveyed in the two of habitat characteristics, at multiple spatial scales forests (74 yr and 112 yr, respectively). A large proportion (Monkk¨ onen¨ et al. 2004, Gotelli et al. 2010, Ricklefs 2013). of upland deciduous stands in both forests represents a The relationship between Canada Warbler occupancy and ‘‘middle age class’’ with high canopy cover that does not species co-occurrence may reflect interspecific (positive and allow for the development of complex subcanopy vegeta- negative) interactions or preferences in habitat characteris- tion. In deciduous stands, this age class appears to be less tics or may be spurious. The positive association between suitable for Canada Warblers than younger age classes that Magnolia Warbler abundance and Canada Warbler occu- are associated with a well-developed understory. It is likely pancy in lowland conifer stands is likely associated with that occupancy has a quadratic relationship with stand age commonalities in habitat use between the 2 species. in deciduous forest stands. The probability of local Magnolia Warblers are often associated with regenerating extinction is lowest (and habitat suitability is highest) in clear-cuts (Titterington et al. 1979) and disturbed conifer early-successional stands (0–30 yr) because of increased stands with increased shrub layers (Burris and Haney 2006), ground cover and shrub layers from harvest, and in mature similar to the preferences of Canada Warblers. Hermit stands (.80 yr) because of gap processes that increase Thrush and Black-and-white Warbler abundances were subcanopy complexity; local extinction is highest (and negatively associated with Canada Warbler occupancy. The habitat suitability lowest) in middle-aged (30–80 yr) mechanisms of these influences are unknown; they do not deciduous stands. We included age as a quadratic term necessarily imply interspecific competition. Interestingly, in in preliminary modeling efforts, but these models had upland conifer stands of Chippewa, Black-and-white War- extensive convergence issues. Survey locations for the bler abundance was the only significant predictor of Minnesota monitoring program were distributed in a occupancy. Our models showed that Canada Warblers and stratified random manner, so our study included relatively Black-and-white Warblers occupied the same stands when fewer early-successional and mature deciduous forest Black-and-white Warbler abundance was low (,1.3 obser- stands than middle-aged stands (Minnesota Forest Re- vations) but were less likely to occupy a stand in years with source Council 2014). Conducting additional surveys that high (.1.9 observations) Black-and-white Warbler abun- focus on early-successional and mature deciduous forest dance. Both of these species nest on or near the ground and stands will be beneficial for assessing long-term dynamics exhibit similar habitat preferences in Chippewa. Although in mature stands in Minnesota. there are no documented interactions between Canada Probability of Canada Warbler occupancy was highest Warblers and Black-and-white Warblers, several sources cite in mixed forest stands. The results reported here suggest aggressive behavior by Black-and-white Warblers. For that small patches of mixed forest stands had negative example, Morse (1970, 1989) reported that ‘‘on territory, effects on species persistence. Area sensitivity in this [Black-and-white Warblers are] inclined to become aggres- species has been well documented in many studies, sive toward other wood-warblers.’’ Similarly, the Hermit especially in the context of fragmentation of forests in an Thrush is a ground-nesting species and may limit available agricultural or urbanized landscape (Ambuel and Temple nest-site locations in certain habitats. The Hermit Thrush is 1983, Robbins et al. 1989). However, the species has been not known to be an aggressive species (Dellinger et al. 2012); reported to be relatively tolerant of habitat fragmentation however, aggressive interactions among thrush species have that results from forest harvesting (Schmiegelow et al. been documented (Able and Noon 1976, Freeman and 1997). Management plans that aim to create or maintain Montgomery 2016). Canada Warblers are among the last large extents of mixed forest stands will likely benefit species to arrive on the breeding grounds, which may make Canada Warblers. them more vulnerable to exclusion by other ground-nesting

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 524 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi species that have already established territories and begun but the long-term population changes may also be linked nesting. Alternatively, Hermit Thrush and Black-and-white with forest management practices and alterations to Warbler abundance may be associated with untested biotic natural disturbance regimes. For example, short rotations or abiotic factors, such as ground-cover substrate, which in of logging (40–70 yr) in relation to historical fire turn influence Canada Warbler occupancy. disturbances (Heinselman 1996) may prevent a high We caution that co-occurrence analysis was conducted proportion of deciduous stands from reaching mature, using raw (unadjusted) species counts; potentially, differ- mixed status. Those natural disturbance regimes likely ences in detectability among species could have biased the provided habitats that contributed to the long-term results. However, Etterson et al. (2009) reported that persistence of Canada Warblers across the landscape. trends calculated from raw counts and detectability- Additional information on the species’ breeding ecology adjusted counts did not differ, given large within-year (e.g., nesting success, postfledging activity, site fidelity) in sample sizes and standardized data-gathering methods, as relation to habitat and landscape context are also needed. were used in the present study. The detection results of Information on nesting success and food availability in Canada Warbler occupancy models may also support an different habitat types across large spatial scales would be an argument for constant detectability. Continued research in important addition to understanding mechanisms associated this area will provide important information for assessing with occupancy dynamics. Importantly, logging disturbance the relative influence of positive, negative, and neutral was not included in final occupancy models, likely because of species co-occurrence on long-term population dynamics. limitations in the availability of precise, consistent distur- bance data throughout the duration of the study. However, in Conclusions the eastern part of the species’ range, recent studies found Our results identify underlying processes associated with that certain logging disturbances positively influenced long-term habitat use across large, dynamic landscapes. Canada Warblers. Given the important effects that forest This information provides insight for conservation and age, percent open cover, and mean patch size had on Canada management efforts to restore Canada Warbler populations Warbler dynamics in the present study, we suggest that a and highlights the importance of including multiple habitat large-scale, long-term experimental approach be used to types and scales in the conservation of species. Occupancy assess the direct effects of harvest on Canada Warblers. of Canada Warblers was highest in mixed forest stands and Extensive changes in habitats and landscapes are occur- lowland conifer stands. Habitat management plans that ring in the central and northern portions of the Canada maximize patch size of mixed aspen stands and minimize Warbler’s breeding range, due to climate change, forest open areas in the matrix surrounding lowland conifer management, and agricultural and residential development stands will likely increase the persistence of this species over (Mladenoff et al. 1997, Wolter and White 2002, Frelich and time. The relationship between consistent occupancy and Reich 2009, Galatowitsch et al. 2009, Wells 2011, Sturtevant high survival and reproductive success has been reported in et al. 2014). These widespread changes signal the need for other species (Lee and Bond 2015, Dugger et al. 2016); urgent conservation action for this species. The present however, these types of data were not collected as part of the study used data from a long-term, large-scale, forest-bird current monitoring program. Future research focused on monitoring program to increase our understanding of nesting success and survival in this study area will aid in Canada Warbler population dynamics in relation to habitat understanding the relationship between occupancy and and landscape context. Our results describe the complexity long-term demography of Canada Warblers and will of the spatial and temporal responses that drive Canada improve future conservation efforts. Warbler habitat occupancy, information that can help us Our results also show that young forests can provide make better forest management decisions and ultimately habitat for Canada Warblers on a short-term basis but that, improve conservation of the species. as deciduous forest stands mature to the midsuccessional age class (31–80 yr old), the stands become less suitable for Canada Warblers. As deciduous stands age beyond ACKNOWLEDGMENTS conventional harvest periods (.75 yr old), stand hetero- geneity increases through gap dynamic processes that This publication is NRRI contribution no. 606. Funding statement: provide suitable long-term Canada Warbler habitat. Our Funding was provided by the U.S. Forest Service, Chippewa and Superior National Forests, UPM results suggest that forest management practices or natural Blandin, the National Council of Air and Stream Improve- disturbances that increase the density of understory ment, Inc., and the Potlatch Corporation. None of our funders vegetation and adequately retain canopy cover are had any influence on the content of the manuscript or beneficial to the species. Reitsma et al. (2010) suggested required approval of the final manuscript to be published. that Canada Warbler population declines likely occur in Ethics statement: No were captured or handled, and response to forest succession. Our results agree with this, no harm was inflicted on any wild population.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 525

Author contributions: A.G. completed all analyses and wrote Ithaca, NY, USA. http://bna.birds.cornell.edu/bna/species/261 the paper. G.N. is the principal investigator for the Minnesota doi:10.2173/bna.261 National Forest Breeding Bird Monitoring Program; he also Diamond, J. M. (1975). Assembly of species communities. In contributed intellectual input, guidance on relevant literature, Ecology and Evolution of Communities (M. L. Cody and J. M. and extensive comments and edits. Diamond, Editors). Harvard University Press, Cambridge, MA, USA. pp. 342–444. LITERATURE CITED Drapeau, P., A. Leduc, J.-F. Giroux, J.-P. L. Savard, Y. Bergeron, and W. L. Vickery (2000). Landscape-scale disturbances and Able, K. P., and B. R. Noon (1976). Avian community structure changes in bird communities of boreal mixed-wood forests. along elevational gradients in the northeastern United States. Ecological Monographs 70:423–444. Oecologia 26:275–294. Dugger, K. M., E. D. Forsman, A. B. Franklin, R. J. Davis, G. C. Ambuel, B., and S. A. Temple (1983). Area-dependent changes in White, C. J. Schwarz, K. P. Burnham, J. D. Nichols, J. E. Hines, C. the bird communities and vegetation of southern Wisconsin B. Yackulic, P. F. Doherty, et al. (2016). The effects of habitat, forests. Ecology 64:1057–1068. climate, and Barred Owls on long-term demography of Becker, D. A., P. B. Wood, and P. D. Keyser (2012). Canada Northern Spotted Owls. The Condor: Ornithological Applica- Warbler use of harvested stands following timber manage- tions 118:57–116. ment in the southern portion of their range. Forest Ecology Enns, K. A., and C. Siddle (1996). The distribution, abundance, and Management 276:1–9. and habitat requirements of selected birds of the Broughton, R. K., R. A. Hill, and S. A. Hinsley (2013). Relationships Boreal and Taiga Plains of British Columbia. Wildlife Working between patterns of habitat cover and the historical Report WR-76. Ministry of Environment, Land and Parks, distribution of the Marsh Tit, Willow Tit and Lesser Spotted Wildlife Branch, Victoria, British Columbia, Canada. Woodpecker in Britain. Ecological Informatics 14:25–30. Enser, R. W. (1992). The atlas of breeding birds in Rhode Island. Burnham, K. P., and D. R. Anderson (2002). Model Selection and Rhode Island Department of Environmental Management, Multimodel Inference: A Practical Information-Theoretic Providence, RI, USA. Approach, second edition. Springer, New York, NY, USA. Environment Canada (2015). Recovery Strategy for Canada Burris, J. M., and A. Haney (2006). Bird communities before and Warbler (Cardellina canadensis) in Canada [Proposed]. Species after a catastrophic blowdown in a Great Lakes pine forest. at Risk Act Recovery Strategy Series. Environment Canada, Bird Populations 7:10–20. Ottawa. Butcher, G. S., D. K. Niven, A. O. Panjabi, D. N. Pashley, and K. V. Etterson, M. A., G. J. Niemi, and N. P. Danz (2009). Estimating the Rosenberg (2007). The 2007 WatchList for United States Birds. effects of detection heterogeneity and overdispersion on American Birds 61:18–25. trends estimated from avian point counts. Ecological Christian, D. P., J. M. Hanowski, M. Reuvers-House, G. J. Niemi, J. Applications 19:2049–2066. G. Blake, and W. E. Berguson (1996). Effects of mechanical Faccio, S. D. (2003). Effects of ice storm-created gaps on forest strip thinning of aspen on small mammals and breeding breeding bird communities in central Vermont. Forest birds in northern Minnesota, U.S.A. Canadian Journal of Ecology and Management 186:133–145. Forest Research 26:1284–1294. Ferraz, G., J. D. Nichols, J. E. Hines, P. C. Stouffer, R. O. Conway, C. J. (1999). Canada Warbler ( canadensis). In Bierregaard, and T. E. Lovejoy (2007). A large-scale defores- The Birds of North America, no. 421 (A. Poole and F. Gill, tation experiment: Effects of patch area and isolation on Editors). Birds of North America, Philadelphia, PA, USA. doi:10. Amazon birds. Science 315:238–241. 2173/bna.421 Freeman, B. G., and G. Montgomery (2016). Interspecific Cooper, J. M., K. A. Enns, and M. G. Shepard (1997). Status of the aggression by the Swainson’s Thrush (Catharus ustulatus) Canada Warbler in British Columbia. Wildlife Working Report may limit the distribution of the threatened Bicknell’s Thrush WR-81. Ministry of Environment, Land and Parks, Wildlife (Catharus bicknelli) in the Adirondack Mountains. The Condor: Branch, Victoria, British Columbia, Canada. Ornithological Applications 118:169–178. COSEWIC (2012). Canadian Wildlife Species at Risk. Committee Freemark, K., and B. Collins (1992). Landscape ecology of birds on the Status of Endangered Wildlife in Canada. http://www. breeding in temperature forest fragments. In Ecology and cosewic.gc.ca/eng/sct0/index_e.cfm Conservation of Neotropical Migrant Landbirds (J. M. Hagan Crawford, H. S., and D. T. Jennings (1989). Predation by birds on III and D. W. Johnston, Editors). Smithsonian Institution Press, spruce budworm Choristoneura fumiferana: Functional, nu- Washington, DC, USA. pp. 443–454. merical, and total responses. Ecology 70:152–163. Frelich, L. E., and P. B. Reich (1995). Spatial patterns and Cumming, S. G., and C. S. Machtans (2001). Stand-level models of succession in a Minnesota southern boreal forest. Ecological the abundances of some songbird species in the Liard Valley, Monographs 65:325–346. North West Territories. Report prepared for the Canadian Frelich, L. E., and P. B. Reich (2009). Will environmental changes Wildlife Service, Yellowknife, Northwest Territories, Canada. reinforce the impact of global warming on the prairie–forest DeGraaf, R. M., J. B. Hestbeck, and M. Yamasaki (1998). border of central North America? Frontiers in Ecology and the Associations between breeding bird abundance and stand Environment 8:371–378. structure in the White Mountains, New Hampshire and Gainsbury, A. M., and G. R. Colli (2003). Lizard assemblages from Maine, USA. Forest Ecology and Management 103:217–233. natural cerrado enclaves in southwestern Amazonia: The role Dellinger, R., P. B. Wood, P. W. Jones, and T. M. Donovan (2012). of stochastic extinctions and isolation. Biotropica 35:503–519. Hermit Thrush (Catharus guttatus). In Birds of North America Galatowitsch, S., L. Frelich, and L. Phillips-Mao (2009). Regional Online, no. 261 (A. Poole, Editor). Cornell Lab of Ornithology, climate change adaptation strategies for biodiversity conser-

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 526 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi

vation in a midcontinental region of North America. global maps of 21st-century forest cover change. Science Biological Conservation 142:2012–2022. 342:850–853. Gates, S., and P. F. Donald (2000). Local extinction of British Heinselman, M. L. (1996). The Boundary Waters Wilderness farmland birds and the prediction of further loss. Journal of Ecosystem. University of Minnesota Press, Minneapolis, MN, Applied Ecology 37:806–820. USA. Goodnow, M. L., and L. R. Reitsma (2011). Nest-site selection in Hines, J. E. (2006). PRESENCE: Software to estimate patch the Canada Warbler (Wilsonia canadensis) in central New occupancy and related parameters. Version 7.3. http://www. Hampshire. Canadian Journal of Zoology 89:1172–1177. mbr-pwrc.usgs.gov/software/presence.html Gotelli, N. J. (2001). Research frontiers in null model analysis. Hobson, K. A., and E. Bayne (2000a). Breeding bird communities Global Ecology & Biogeography 10:337–343. in the boreal forest of western Canada: Consequences of Gotelli, N. J., and A. M. Ellison (2013). EcoSimR 1.00. https://www. ‘‘unmixing’’ the mixedwoods. The Condor 102:759–769. uvm.edu/~ngotelli/EcoSim/EcoSim.html Hobson, K. A., and E. Bayne (2000b). The effects of stand age on Gotelli, N. J., and G. L. Entsminger (2005). EcoSim: Null models avian communities in aspen-dominated forests of central software for ecology. Version 7.72. Acquired Intelligence, Saskatchewan, Canada. Forest Ecology and Management Jericho, VT, USA. 136:121–134. Gotelli, N. J., G. R. Graves, and C. Rahbek (2010). Macroecological Hobson, K. A., D. A. Kirk, and A. R. Smith (2000). A multivariate signals of species interactions in the Danish avifauna. analysis of breeding bird species of western and central Proceedings of the National Academy of Sciences USA 107: Canadian boreal forests: Stand and spatial effects. Ecoscience´ 5030–5035. 7:267–280. Green, J. C. (1995). Birds and Forests. Minnesota Department of Hobson, K. A., and J. Schieck (1999). Changes in bird Natural Resources, St. Paul, MN, USA. communities in boreal mixedwood forest: Harvest and Guillera-Arroita, G. (2011). Impact of sampling with replacement wildfire effects over 30 years. Ecological Applications 9:849– in occupancy studies with spatial replication. Methods in 863. Ecology and Evolution 2:401–406. Howe, R. W., G. J. Niemi, G. J. Lewis, and D. A. Welsh (1997). A Hagan, J. M., and S. L. Grove (1999). Bird abundance and standard method for monitoring songbird populations in the distribution in managed and old-growth forest in Maine. Great Lakes region. Passenger Pigeon 59:182–194. Report MM-9901. Manomet Center for Conservation Scienc- Kendall, W. L., and G. C. White (2009). A cautionary note on es, Brunswick, ME, USA. substituting spatial subunits for repeated temporal sampling Hagan, J. M., P. S. McKinley, A. L. Meehan, and S. L. Grove (1997). in studies of site occupancy. Journal of Applied Ecology 46: Diversity and abundance of landbirds in a northeastern 1182–1188. industrial forest. Journal of Wildlife Management 61:718–735. Lambert, J. D., and S. D. Faccio (2005). Canada Warbler Hallworth, M., P. M. Benham, J. D. Lambert, and L. Reitsma population status, habitat use and stewardship guidelines (2008a). Canada Warbler (Wilsonia canadensis)breeding for the northeastern forests. VINS Technical Report 05-4. ecology in young forest stands compared to a red maple Vermont Institute of Natural Sciences, Woodstock, VT, (Acer rubrum) swamp. Forest Ecology and Management 255: USA. 1353–1358. Lapin, C. N., M. A. Etterson, and G. J. Niemi (2013). Occurrence of Hallworth, M., A. Ueland, E. Anderson, J. Lambert, and L. Reitsma Connecticut Warbler increases with coniferous forest patch (2008b). Habitat selection and site fidelity of Canada Warblers size. The Condor 115:168–177. (Wilsonia canadensis) in central New Hampshire. The Auk 125: Lee, D. E., and M. L. Bond (2015). Occupancy of California 880–888. Spotted Owl sites following a large fire in the Sierra Nevada, Hannon, S. J., S. E. Cotterill, and F. K. A. Schmiegelow (2004). California. The Condor: Ornithological Applications 117:228– Identifying rare species of in managed forests: 236. Application of an ecoregional template to a boreal mixed- Loss, S. R., T. Will, S. S. Loss, and P. P. Marra (2014). Bird-building wood system. Forest Ecology and Management 191:157–170. collisions in the United States: Estimates of annual mortality Hanowski, J., J. Lind, N. Danz, G. Niemi, and T. Jones (2005). and species vulnerability. The Condor: Ornithological Appli- Regional breeding bird monitoring in western Great Lakes cations 116:8–23. national forests. In Bird Conservation Implementation and Lovette, I. J., and W. M. Hochachka (2006). Simultaneous effects Integration in the Americas: Proceedings of the Third of phylogenetic niche conservatism and competition on International Partners in Flight Conference (C. J. Ralph and avian community structure. Ecology 87:S14–S28. T. D. Rich, Editors). USDA Forest Service General Technical MacArthur, R. H. (1958). Population ecology of some warblers of Report PSW-GTR-191. northeastern coniferous forests. Ecology 39:599–619. Hanowski, J. M., and G. J. Niemi (1995). Experimental design Machtans, C. S., and P. B. Latour (2003). Boreal forest songbird considerations for establishing an off-road, habitat-specific communities of the Liard Valley, Northwest Territories, bird monitoring program using point counts. In Monitoring Canada. The Condor 105:27–44. Bird Populations by Point Counts (C. J. Ralph, J. R. Sauer, and Machtans, C. S., and W. E. Thogmartin (2014). Understanding the S. Droege Editors). USDA Forest Service General Technical value of imperfect science from national estimates of bird Report PSW-GTR-149, Pacific Southwest Research Station, mortality from window collisions. The Condor: Ornithological Albany, CA, USA. Applications 116:3–7. Hansen, M. C., P. V. Potapov, R. Moore, M. Hancher, S. A. MacKenzie, D. I., and L. L. Bailey (2004) Assessing the fit of site- Turubanova, A. Tyukavina, D. Thau, S. V. Stehman, S. J. Goetz, occupancy models. Journal of Agricultural, Biological, and T. R. Loveland, A. Kommareddy, et al. (2013). High-resolution Environmental Statistics 9:300–318.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 527

MacKenzie, D. I., J. D. Nichols, G. Lachman, S. Droege, J. A. Royle, Reitsma, L., M. T. Hallworth, and P. M. Benham (2008). Does age and C. A. Langtimm (2002). Estimating site occupancy rates influence territory size, habitat selection, and reproductive when detection probabilities are less than one. Ecology 83: success of male Canada Warblers in central New Hampshire? 2248–2255. Wilson Journal of Ornithology 120:446–459. MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, Rempel, R. S., D. Kaukinen, and A. P. Carr (2012). Patch Analyst and J. E. Hines (2006). Occupancy Estimation and Modeling: and Patch Grid. Ontario Ministry of Natural Resources. Centre Inferring Patterns and Dynamics of Species Occurrence. for Northern Forest Ecosystem Research, Thunder Bay, Academic Press, New York, NY, USA. Ontario, Canada. McGarigal, K., and B. J. Marks (1995). FRAGSTATS: Spatial pattern Ricklefs, R. E. (2013). Habitat-independent spatial structure in analysis program for quantifying landscape structure. Refer- populations of some forest birds in eastern North America. ence manual. Forest Science Department, Oregon State Journal of Animal Ecology 82:145–154. University, Corvallis, OR, USA. Robbins, C. S., D. K. Dawson, and B. A. Dowell (1989). Habitat Minnesota Forest Resource Council (2014). Northeast Landscape area requirements of breeding forest birds of the Middle Forest Resources Plan. Document LP0914. Minnesota Forest Atlantic states. Wildlife Monographs 103. Resource Council, St. Paul, MN, USA. Sauer, J. R., J. E. Hines, J. E. Fallon, K. L. Pardieck, D. J. Ziolkowski, Mitchell, J. M. (1999). Habitat relationships of five northern bird Jr., and W. A. Link (2014). The North American Breeding Bird species breeding in hemlock ravines in Ohio, USA. Natural Survey, Results and Analysis 1966–2012. Version 02.19.2014. Areas Journal 19:3–11. USGS Patuxent Wildlife Research Center, Laurel, MD. Mladenoff, D., G. J. Niemi, and M. A. White (1997). Effects of Schieck, J., and K. A. Hobson (2000). Bird communities associated changing landscape pattern and U.S.G.S. land cover data with live residual tree patches within cut blocks and burned variability on ecoregion discrimination across a forest– habitat in mixedwood boreal forests. Canadian Journal of agriculture gradient. Landscape Ecology 12:379–396. Forest Research 30:1281–1295. Monkk¨ onen,¨ M., J. T. Forsman, and R. L. Thomson (2004). Schieck, J., M. Nietfeid, and J. B. Stelfox (1995). Differences in bird Qualitative geographical variation in interspecific interac- species richness and abundance among three successional tions. Ecography 27:112–118. stages of aspen-dominated boreal forests. Canadian Journal Morse, D. H. (1970). Ecological aspects of some mixed-species of Zoology 73:1417–1431. foraging flocks of birds. Ecological Monographs 40:119–168. Schieck, J., K. Stuart-Smith, and M. R. Norton (2000). Bird Morse, D. H. (1989). American Warblers: An Ecological and communities are affected by amount and dispersion of Behavioral Perspective. Harvard University Press, Cambridge, vegetation retained in mixedwood boreal forest harvest MA, USA. areas. Forest Ecology and Management 126:239–254. Niemi, G., J. Hanowski, P. Helle, R. Howe, M. Monkk¨ onen,¨ L. Schmiegelow, F., E. Bayne, S. Cumming, S. Song, T. Fontaine, S. Venier, and D. Welsh (1998). Ecological sustainability of birds Hache,´ L. Mahon, P. Solymos,´ N. Barker, S. Matsuoka, and D. in boreal forests. Conservation Ecology 2:17. Stralberg (2014). Boreal Avian Modelling Project: Predictive Niemi, G. J., R. W. Howe, B. R. Sturtevant, L. R. Parker, A. R. Grinde, tools for the monitoring and assessment of boreal birds in N. P. Danz, M. Nelson, E. J. Zlonis, N. G. Walton, E. E. Giese, Canada. 2013–2014 Annual Report to Environment Canada. and S. M. Lietz (2016). Analysis of long-term forest bird Schmiegelow, F. K. A., C. S. Machtans, and S. J. Hannon (1997). Are monitoring in national forests of the western Great Lakes boreal birds resilient to forest fragmentation? An experimental region. USDA Forest Service General Technical Report. In study of short-term community responses. Ecology 78:1914–1932. press. Skally, C. (2000). Integration of forest inventory GIS data in Niven, D. K., G. S. Butcher, and G. T. Bancroft (2009). Christmas Minnesota. Minnesota Forest Resources Council, Minnesota bird counts and climate change: Northward shifts in early Interagency Information Cooperative, St. Paul, MN, USA. winter abundance. American Birds 63:10–15. Sodhi, N. S., and C. A. Paszkowski (1995). Habitat use and Pattison, R. R., C. M. D’Antonio, T. Dudley, K. K. Allander, and B. foraging behavior of four parulid warblers in a second- Rice (2011). Early impacts of biological control on canopy growth forest. Journal of Field Ornithology 66:277–288. cover and water use of the invasive saltcedar tree (Tamarix Stone, L., and A. Roberts (1990). The checkerboard score and spp.) in western Nevada, USA. Oecologia 165:605–616. species distributions. Oecologia 85:74–79. Pianka, E. R. (1986). Ecology and Natural History of Desert Streby, H. M., J. P. Loegering, and D. E. Andersen (2012). Spot- Lizards. Princeton University Press, Princeton, NJ, USA. mapping underestimates song-territory size and use of Porter, W. F., and M. A. Jarzyna (2013). Effects of landscape-scale mature forest by breeding male Golden-winged Warblers in forest change on the range contraction of Ruffed Grouse in Minnesota, USA. Wildlife Society Bulletin 36:40–46. New York State, USA. Wildlife Society Bulletin 37:198–208. Sturtevant, B. R., B. R. Miranda, P. T. Wolter, P. M. A. James, M.-J. Ralph, C. J., J. R. Sauer, and S. Droege (Editors) (1995). Monitoring Fortin, and P. A. Townsend (2014). Forest recovery patterns in bird populations by point counts. USDA Forest Service response to divergent disturbance regimes in the Border General Technical Report PSW-GTR-149. Volume 2. Pacific Lakes region of Minnesota (USA) and Ontario (Canada). Southwest Research Station, Albany, CA, USA. Forest Ecology and Management 331:199–211. Reitsma, L., M. Goodnow, M. T. Hallworth, and C. J. Conway Titterington, R. W., H. S. Crawford, and B. N. Burgason (1979). (2010). Canada Warbler (Cardellina canadensis). In Birds of Songbird responses to commercial clear-cutting in Maine North America Online (A. Poole, Editor). Cornell Lab of spruce–fir forests. Journal of Wildlife Management 42:602–609. Ornithology, Ithaca, NY, USA. http://bna.birds.cornell.edu/ Ulrich, W., M. Almeida-Neto, and N. J. Gotelli (2009). A bna/species/421 consumer’s guide to nestedness analysis. Oikos 118:3–17.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 528 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi

Ulrich, W., and N. J. Gotelli (2007a). Disentangling community Wells, J. V. (Editor) (2011). Boreal birds of North America. Studies patterns of nestedness and species co-occurrence. Oikos 116: in Avian Biology 41. 2053–2061. Wolter, P., and M. White (2002). Recent forest cover type Ulrich, W., and N. J. Gotelli (2007b). Null model analysis of transitions and landscape structural changes in northeast species nestedness patterns. Ecology 88:1824–1831. Minnesota, USA. Landscape Ecology 17:133–155. U.S. Fish and Wildlife Service (2008). Birds of Conservation Yntze,H.,A.M.Wilson,R.Renfrew,J.Walsh,P.G.Rodewald,J. Concern 2008. U.S. Department of Interior, Fish and Wildlife Baldy, and L. L. Manne (2015). Regional variability in Service, Division of Migratory Bird Management, Arlington, extinction thresholds for forest birds in the north-eastern VA, USA. United States: An examination of potential drivers using U.S. Geological Survey (2011). National Land Cover, version 2. long-term breeding bird atlas datasets. Diversity and U.S. Department of the Interior, USGS Gap Analysis Program. Distributions 21:686–697. Webb, W. L., D. F. Behrend, and B. Saisorn (1977). Effect of Zlonis, E. J., and G. J. Niemi (2014). Avian communities of logging on songbird populations in a northern hardwood managed and wilderness hemiboreal forests. Forest Ecology forest. Wildlife Monographs 55. and Management 328:26–34.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 529

APPENDIX TABLE 1. Candidate models in forest dynamics analysis used to build final models to assess Canada Warbler occupancy dynamics in Chippewa and Superior National Forests from 1995 to 2014. Models were built using stand-characteristic covariates and land-cover covariates that were calculated at the 100 m and 500 m buffer scales. Based on the results of the preliminary analysis, detectability (p) was modeled as constant in Chippewa and was modeled with annual variation in Superior. Local Predicted Candidate model Colonization extinction KChippewa KSuperior effect Null c(.) e(.) 4 4 . c(.) e(.) 23 23 . Annual variation c(year) e(year) 40 59 . 100 m buffer Stand-characteristic variables Age a c(age) e(.) 4 23 þ c(.) e(age) 4 23 Ground cover a c(grcov) e(.) 4 23 þ c(.) e(grcov) 4 23 Shrub density a c(shrub) e(.) 4 23 þ c(.) e(shrub) 4 23 Distance to road c(disrd) e(.) 4 23 þ c(.) e(disrd) 4 23 Logging disturbance (500 m) c(logdist100) e(.) 4 23 þ c(.) e(logdist100) 4 23 Mean patch size (100 m) a c(mps100) e(.) 4 23 þ c(.) e(mps100) 4 23 Total edge (100 m) c(te100) e(.) 4 23 c(.) e(te100) 4 23 þ Land-cover variables Upland conifer (100 m) a c(upcon100) e(.) 4 23 . c(.) e(upcon100) 4 23 . Upland deciduous (100 m) c(upcon100) e(.) 4 23 . c(.) e(upcon100) 4 23 . Upland mixed forest (100 m) a c(upmix100) e(.) 4 23 . c(.) e(upmix100) 4 23 . Lowland conifer (100 m) a c(llcon100) e(.) 4 23 . c(.) e(llcon100) 4 23 . Aquatic (%) (100 m) a c(wet100) e(.) 4 23 þ c(.) e(wet100) 4 23 Open (%) (100 m) a c(open100) e(.) 4 23 c(.) e(open100) 4 23 þ 500 m buffer Stand-characteristic variables Logging disturbance (500 m) c(logdist500) e(.) 4 23 þ c(.) e(logdist500) 4 23 Mean patch size (500 m) a c(mps500) e(.) 4 23 þ c(.) e(mps500) 4 23 Total edge (500 m) c(te500) e(.) 4 23 c(.) e(te500) 4 23 þ Land-cover variables Upland conifer (500 m) a c(upcon500) e(.) 4 23 . c(.) e(upcon500) 4 23 . Upland deciduous (500 m) c(upcon500) e(.) 4 23 . c(.) e(upcon500) 4 23 . Upland mixed forest (500 m) a c(upmix500) e(.) 4 23 . c(.) e(upmix500) 4 23 . Lowland conifer (500 m) a c(llcon500) e(.) 4 23 . c(.) e(llcon500) 4 23 . Aquatic (%) (500 m) a c(wet) e(.) 4 23 þ c(.) e(wet) 4 23 Open (%) (500 m) a c(open500) e(.) 4 23 c(.) e(open500) 4 23 þ a Natural log transformation was used prior to analysis.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society 530 Canada Warbler occupancy dynamics A. R. Grinde and G. J. Niemi

APPENDIX TABLE 2. Candidate models in cover-type dynamics analysis used to build final models to assess Canada Warbler occupancy dynamics in Chippewa and Superior National Forests from 1995 to 2014. Models were built using stand characteristics and species co-occurrence covariates. Species co-occurrence covariates were included on the basis of co-occurrence analysis. Based on the results of the preliminary analysis, detectability (p) was modeled as constant in both forests and in all forest-cover types. Upland deciduous, upland conifer, and lowland conifer cover types were modeled in both forests; mixed forest-cover type was modeled in Superior only. Scientific names of species are given in Appendix Table 3. Local Predicted Candidate model Occupancy extinction Detection K effect

Null w(.) e(.) p(.) 3 . Annual variation w(.) e(year) p(.) 21 . w(year) e(.) p(.) 22 . Stand-characteristic variables Age a w(age) e(.) p(.) 3 þ w(.) e(age) p(.) 3 Ground cover a w(grcov) e(.) p(.) 3 þ w(.) e(grcov) p(.) 3 Shrub density a w(shrub) e(.) p(.) 3 þ w(.) e(shrub) p(.) 3 Distance to road w(disrd) e(.) p(.) 3 þ w(.) e(disrd) p(.) 3 Aquatic (%) (500 m) a w(wet) e(.) p(.) 3 þ w(.) e(wet) p(.) 3 Open (%) (500 m) a w(open500) e(.) p(.) 3 w(.) e(open500) p(.) 3 þ Mean patch size (100 m) a w(mps100) e(.) p(.) 3 þ w(.) e(mps100) p(.) 3 Mean patch size (500 m) a w(mps500) e(.) p(.) 3 þ w(.) e(mps500) p(.) 3 Total edge (100 m) w(te100) e(.) p(.) 3 w(.) e(te100) p(.) 3 þ Total edge (500 m) w(te500) e(.) p(.) 3 w(.) e(te500) p(.) 3 þ Species co-occurrence variables Black-and-white Warbler a w(BAWW) e(.) p(.) 3 . Blackburnian Warbler a w(BLBW) e(.) p(.) 3 . Black-throated Green Warbler a w(BTNW) e(.) p(.) 3 . Hermit Thrush a w(HETH) e(.) p(.) 3 . Magnolia Warbler a w(MAWA) e(.) p(.) 3 . Mourning Warbler a w(MOWA) e(.) p(.) 3 . a Natural log transformation was used prior to analysis.

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society A. R. Grinde and G. J. Niemi Canada Warbler occupancy dynamics 531

APPENDIX TABLE 3. Bird species and C-scores (see text) from tests for nonrandom patterns of co-occurrence with Canada Warblers in Chippewa and Superior National Forests. The C-score index was used as a quantitative co-occurrence index to determine the probability of all species’ co-occurrence with Canada Warblers. C-scores were standardized using the mean of simulated indices for comparison between forest-cover types. Species in bold are those that had high (ranked in top 5) checkerboard-unit values with Canada Warblers in 3 cover types and were used as covariates in occupancy models. Forest-cover type Bird species C-score Chippewa Upland deciduous Black-and-white Warbler (Mniotilta varia) 3.21 Mourning Warbler (Geothlypis philadelphia) 2.75 Red-breasted Nuthatch (Sitta canadensis) 2.75 Swamp Sparrow (Melospiza georgiana) 2.48 Magnolia Warbler ( magnolia) 2.06 Upland mixed Ovenbird (Seiurus aurocapilla) 6.67 Black-and-white Warbler 4.00 Mourning Warbler 4.00 Hermit Thrush (Catharus guttatus) 3.33 Blackburnian Warbler (Setophaga fusca) 2.67 Upland conifer Black-throated Green Warbler (Setophaga virens) 5.07 Blackburnian Warbler 3.16 Brown Creeper (Certhia americana) 2.63 Hermit Thrush 2.37 Magnolia Warbler 2.30 Lowland conifer Black-and-white Warbler 5.88 Black-capped Chickadee (Poecile atricapillus) 3.78 Ruby-throated Hummingbird (Archilochus colubris) 3.78 Brown Creeper 3.78 Song Sparrow (Melospiza melodia) 2.94 Superior Upland deciduous Black-throated Green Warbler 2.13 Blackburnian Warbler 1.93 American Robin (Turdus migratorius) 1.82 Mourning Warbler 1.64 Hermit Thrush 1.47 Upland mixed Chestnut-sided Warbler (Setophaga pensylvanica) 2.63 Winter Wren (Troglodytes hiemalis) 2.26 Blackburnian Warbler 1.64 Chipping Sparrow (Spizella passerina) 1.50 Hermit Thrush 1.50 Upland conifer Magnolia Warbler 3.16 Nashville Warbler (Oreothlypis ruficapilla) 3.16 Least Flycatcher (Empidonax minimus) 2.84 Chestnut-sided Warbler 2.53 Pileated Woodpecker (Dryocopus pileatus) 2.53 Lowland conifer Ruffed Grouse (Bonasa umbellus) 4.74 Song Sparrow 4.15 Eastern Wood-Pewee (Contopus virens) 3.16 Winter Wren 2.17 Black-throated Green Warbler 1.98

The Condor: Ornithological Applications 118:513–531, Q 2016 Cooper Ornithological Society